Smart Meter Data Analytics for Load Prediction using Extreme Learning Machines and Artificial Neural Networks

Load forecasting is one of the essential component in Smart Grid operation as it plays a significant role in making decisions such as planning, generation scheduling and operation, electricity pricing, customer satisfaction and system security. The date old load forecasting problem has received renewed efforts due to the arrival of Smart Meters in the Smart Grid system. Residential load forecasting is a relatively new area focusing the customer load prediction. Unlike the aggregate level load forecasting, residential load forecasting is difficult due to the high variable nature of residential load curves. A comparison of load forecasting algorithm using Extreme Learning Machines(ELM) with the traditional approach using Neural Networks (NN) is presented in this paper. Smart meter data obtained from a residential customer is used for testing the performance of the proposed hybrid algorithm. The accuracy and efficiency of the model is analyzed using statistical parameters like MAE, MAPE and RMSE. Based on the experimental results ELM based load forecasting gives better accuracy when compared to the traditional NN based algorithm.